Overview
How might we design smart recommendations so people can trust and act on AI output?
When to use
- Essential for e-commerce platforms, content discovery applications, and marketplaces where personalized recommendations drive engagement and sales.
When to skip
- Cold-start users with no signal beyond generic bestsellers.
- Regulated contexts where personalized pricing or health advice is restricted.
- Recommendations that duplicate search results with no added value.
Rules
“Recommended for you” with no explainability or dismiss.
Same block on every page regardless of context.
Boosting paid placements without disclosure.
Recommendations that fight the user’s stated filters.
Evidence
| Product | Implementation |
|---|---|
| Amazon | “Customers also bought” and session-aware carousels. |
| Netflix | Row rankings from taste and watch history. |
| Spotify | Discover Weekly and contextual mixes. |
| Shopify | Merchant recommendation apps on product and cart pages. |
